海底视像图像的灰度不均匀校正方法的研究

Removing Non-uniform Illumination of Deep_sea Floor Pictures

点光源照射的亮度不均匀使获得的海底视像图像的灰度分布极不均匀 ,给海底多金属结核矿物的颗粒自动识别、粒径测量、储量的计算带来误差。本文针对点光源照射的多金属结核海底视像图像 ,提出了 3种方法进行灰度不均匀校正 ,并作了比较。实验结果表明 ,统计原图灰度值的二维分布特性 ,用二阶三次多项式函数拟合此灰度分布 ,再逐像元进行灰度位伸的方法能更有效地消除点光源对海底视像图像灰度的影响。

There exist abundant multi_metal grains in over 5 000 m deep sea floor.In order to exploit the deep_sea floor mineral,we have taken some deep_sea floor pictures for study.Fig 1 shows a typical picture.There are dark metal grains pixels (the target objects),and grayer sands pixels (the background objects).Point light source is necessarily used when we take pictures for it is very dark in deep_sea floor.However,it causes serious non_uniform illumination in these pictures,which enlarges errors when measuring grain's radius and calculating mental conservation. [FK(W27。40ZQ] Three simple methods are developed in this paper for removing the non_uniform exposure effects.(1) Local enhacement and filtering.By setting appropriate parameters in the filter,we can force the brightness and contrastness of pixels in a local window into appointed target mean grey and contrast value.As a result,the brightness and contrastness will be improved in insufficient exposure area,partly reducing non_illumination effects.(2)Substracting the source image from the background image,which merely consist of sands pixels by extracting the local maximum grey value.It is not easy to decide the local window size to make it appropriate for not only being large enough to contain at least one pixel of metal grain and one pixel of sands but also being small enough to make the intensity of the point light source distributed approximately equivalent inside this local window.Generally,it changes with the size and density of metal grains in digital image.After image substracting,most of the effects of non_uniform illumination have been removed from the deep_sea floor pictures.(3)Statistical fitting grey distribution and stretching.First we calculate the local maximum and minimum grey value all over the image using the above second method,then fit two planar cubic equations with these statistical data,and then based on the fitted data all over the image,stretch the gray value pixel by pixel. Theoretically,the third method integrates the advantages of the second one,and can completely remove the effects of non_illumination from the image.For analyzing and comparing the transformation of grey value before and after each processing step,we give the trendlines of grey value of pixels in line 263,which reflects the one_dimentional grey and contrastness distribution characteristic in these pictures.From the trendlines,we can draw a conclution that the last approach is more effective than the other two in removing point light source exposure effect from deep_sea floor pictures.In a word,though we do not make use of the ideal model of point light source in the whole experiment,we have also solved the problems.The approaches employed here demonstrate the effectiveness and practical significance.